APPLIED PHYSICS LETTERS
VOLUME 81, NUMBER 27
30 DECEMBER 2002
Resonance-enhanced laser-induced plasma spectroscopy for sensitive elemental analysis: Elucidation of enhancement mechanisms S. L. Lui and N. H. Cheunga) Department of Physics, Hong Kong Baptist University, Kowloon Tong, Hong Kong, China
共Received 4 September 2002; accepted 30 October 2002兲 When performing laser-induced plasma spectroscopy for elemental analysis, the analyte signal-to-noise ratio increased from four to over fifty if the plume was reheated by a dye laser pulse tuned to resonant absorption. Time-resolved studies showed that the enhancement was not due to resonance photoionization. Rather, efficient and controlled rekindling of a larger plume volume was the key mechanism. The signal-to-noise ratio further increased to over a hundred if the atmosphere was replaced by a low-pressure heavy inert gas. The ambient gas helped confine and thermally insulate the expanding vapor. © 2002 American Institute of Physics. 关DOI: 10.1063/1.1532774兴
Laser-induced plasma spectroscopy 共also known as laser-induced breakdown spectroscopy兲 is a versatile technique for elemental analysis. It can be performed remotely on practically any sample, and applications ranging from space exploration to art authentication have been demonstrated.1–3 Unfortunately, laser ignition of plasmas is chaotic and violent. Reproducibility and sensitivity are, therefore, compromised. We recently showed that the sensitivity could be improved if the plume was photoexcited resonantly.4 For example, we laser ablated a potassium-rich target doped with sodium as a test analyte. The expanding plume was then intercepted by a dye laser pulse 共404 nm兲 to resonantly excite the K atoms (4 2 S 1/2→5 2 P 3/2). The Na 589 nm emissions were enhanced. Undoubtedly, the analytical performance of resonance-enhanced laser-induced plasma spectroscopy 共RELIPS兲 can be improved if the underlying processes are better understood. In this letter, we report a study of the RELIPS mechanism and illustrate how the sensitivity could be further enhanced. Cylindrical pellets of potassium iodate (KIO3 ) containing 35 ppm of sodium were used as targets 共for details, see Ref. 4兲. Lithium 共55 ppm兲 was added when the plasma temperature T and electron density n e were to be measured:5 T was determined from the intensity ratio of the Li 610.3 and 670.8 nm lines, which was shown to be consistent with that deduced from the more conventional iron lines.6 n e was determined from the Stark width of the Li 610.3 nm line, which was shown to agree with estimates based on the linewidths of the H␣ and H transitions.6,7 However, n e deduced from the width of the Li 670.8 nm line, was found to be overestimated, and corrections were necessary. In separate experiments, pellets of sodium bicarbonate (NaHCO3 ) doped with 50 ppm lithium were used as targets. The experimental setup, shown schematically in Fig. 1, was similar to the one used previously.4,8 Briefly, a laser pulse 共532 nm, 10 ns, and 10 Hz兲 from a Nd:YAG laser was apertured and imaged onto the side of the rotating target. 30 ns later, a dye laser pulse 共9 ns and 10 Hz兲 was focused normally onto the same spot. Because near-threshold ablaa兲
Author to whom correspondence should be addressed; electronic mail:
[email protected]
tions were studied, the fluence of the 532 nm laser pulse was only about 800 mJ cm⫺2 共Ref. 9兲. The fluence of the dye laser pulse was about 460 mJ cm⫺2, which was marginally ablative. Unless stated otherwise, ablations were done under 1 atm of air although the sample chamber could be filled with different gases to any pressure. Light emissions from the plasma were imaged axially onto the entrance slit of a spectrograph equipped with an intensified charge-coupled device 共ICCD兲. Analyte spectra were captured using a slit width of 300 m, giving a 0.2 nm resolution. Lithium linewidths were measured with a 100 m slit at 0.08 nm resolution. Timeintegrated spectra were taken with an intensifier gate delay t d of 40 ns 共relative to the onset of the 532 nm pulse兲 and a 5 s gate width t b . For time-resolved studies, t b was 50 ns while t d was scanned. The 50 ns t b was the electronic pulse width. The actual optical gate width was narrower.10 It was characterized by scanning the gate across the 532 nm light pulse. The resultant pulse width was about 37 ns 关full width at half maximum 共FWHM兲兴, and the occurrence of the pulse maximum was located to within ⫾ 5 ns. For that reason, t d was stepped at 5 ns for time-resolved studies. In all cases,
FIG. 1. Schematic diagram of RELIPS setup. A rotating cylindrical pellet of KIO3 housed in a sample cell was ablated by the second-harmonic 共532 nm兲 output of a Nd:YAG laser pulse of 10 ns width. 30 ns later, the expanding plume was intercepted by a dye laser pulse of 9 ns duration and 0.3 nm linewidth centering on 404.4 nm. The plume emissions were directed onto the entrance slit of a spectrograph equipped with an intensified array detector. The sample cell could be evacuated or filled with various ambient gases to any desired pressure.
0003-6951/2002/81(27)/5114/3/$19.00 5114 © 2002 American Institute of Physics Downloaded 27 Dec 2002 to 158.182.14.135. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
Appl. Phys. Lett., Vol. 81, No. 27, 30 December 2002
S. L. Lui and N. H. Cheung
5115
FIG. 2. Plume emission spectra generated by the RELIPS scheme. KIO3 pellets containing 35 ppm of Na were ablated in air by a 532 nm laser pulse at a fluence of about 800 mJ cm⫺2. 30 ns later, the expanding plume was intercepted by a minimally ablative dye laser pulse of about 460 mJ cm⫺2. The plume emissions were time integrated for 5 s starting 10 ns after the firing of the dye laser. Emission spectra produced by 共a兲 the dye laser alone, 共b兲 the 532 nm laser alone, 共c兲 the 532 nm laser followed by an offresonance 407 nm dye laser pulse, 共d兲 similar to 共c兲 except with an onresonance 404 nm dye laser pulse, and 共e兲 similar to 共d兲 except the ambient gas was 13 mbar xenon instead of open air. Since the edge pixels of the charge coupled device 共CCD兲 were not intensified, the spectral trace at the edges served conveniently as baselines in all cases. All spectra were offset vertically for clarity. The SNR for the various traces were about 共a兲 1, 共b兲 4, 共c兲 5, 共d兲 53, and 共e兲 110.
300 events were accumulated before the spectrum was stored and processed. The effectiveness of RELIPS is shown in Fig. 2. Trace 共a兲 was generated with the 404 nm dye laser alone. The edge pixels were not intensified so the two edges served as baselines. Trace 共b兲 was generated with the 532 nm pulse alone. Trace 共c兲 was generated with the 532 nm pulse followed by an off-resonance 共407 nm兲 dye laser pulse. Trace 共d兲 was similar to 共c兲 except with the dye laser tuned to resonance 共404 nm兲. Enhancement is clearly demonstrated. The enhancement may be quantified in terms of the signal-to-noise ratio 共SNR兲. If we define the Na signal as the average intensity under the doublet minus the average background intensity, and noise as the standard deviation of the background, then the SNR of traces 共a兲 through 共d兲 are, respectively, 1, 4, 5, and 53. Trace 共e兲 will be discussed in a later section. The enhancement mechanism was elucidated by capturing time-resolved spectra. They were taken under conditions identical to that of Fig. 2 共traces c and d兲 except with a 50 ns gate width and an air pressure of 350 mbar. A lower pressure was selected to ensure a reasonable signal even under nonresonant conditions 共see next兲. The signal and background intensities, as defined earlier, are plotted against time in Figs. 3共a兲 and 3共b兲, respectively. The dye laser was tuned to either 404 nm for the on-resonance traces 共open circles兲 or 407 nm for the off-resonance traces 共crosses兲. The time axis t is defined as t d ⫹t b /2. It marks the center of the time window. Events earlier than 25 ns were not captured because of overwhelming continuum backgrounds. As can be seen, relative to the off-resonance case, 404 nm light induced a small increase in background but a marked enhancement in signal intensity and lifetime. The evolution of the electron density n e (t) and plasma temperature T(t) is shown in Figs. 3共c兲 and 3共d兲 共darker symbols兲, respectively. Again, resonant effects are very ap-
FIG. 3. Time-resolved plume emissions generated under conditions identical to that of traces 共c兲 and 共d兲 in Fig. 2, except the ambient gas was 350 mbar air and the ICCD gate width was 50 ns. The effective optical gate width was about 37 ns. The time axis t was the ICCD gate delay t d plus half of the gate width t b . Time t was, therefore, the center of the time window. Both onresonance 共circles兲 and off-resonance 共crosses兲 behaviors were shown for easy comparison. The four panels are: 共a兲 sodium signal, 共b兲 continuum background, 共c兲 electron density determined from Stark widths of the Li 610.3 and 670.8 nm lines, and 共d兲 plasma temperature determined from the intensity ratio of the same pair of Li lines, together with the Li 610.3 nm line intensity 共lighter symbols兲.
parent. The cyclic variation in T for the 404 nm case was probably due to shock waves.11 The off-resonance T was only briefly measurable between 60 and 65 ns, when it was found to be about 0.42 eV. At earlier times, the background masked the weak Li 610 nm line. At later times, T dropped below the measurable threshold of 0.35 eV. In addition to the obvious, three subtle observations may be drawn from Fig. 3 that are no less important. First, n e remained high (1016 and 1017 cm⫺3 ) during much of the signal lifetime, be it on or off resonance. It guaranteed local thermal equilibrium 共LTE兲 for well-defined plasma temperatures.12 An unnecessary resonant boost would only contribute to an unwelcome background. Second, the resonant peaking of n e occurred at t ⫽50 ns 关Fig. 3共c兲兴, which was 20 ns after the 404 nm pulse was fired. This noticeable delay suggests that 1⫹1 photoionization,13 K共 4 2 S 1/2兲 ⫹h 共 404 nm兲 →K共 5 2 P 3/2兲 ,
共1兲
K共 5 2 P 3/2兲 ⫹h 共 404 nm兲 →K⫹ ⫹e ⫺ ,
共2兲
could not be the dominant electron production process. Quenching of the excited K (5p) atoms via electron–atom superelastic collisions would compete with photoionization 共process 2兲.14 Given the high n e , collision deexcitation was
Downloaded 27 Dec 2002 to 158.182.14.135. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
5116
Appl. Phys. Lett., Vol. 81, No. 27, 30 December 2002
more likely. The suprathermal electrons so produced eventually impact ionized the ambient atoms to raise n e . At the present dye laser irradiance (⬃107 W cm⫺2 ) and electron density (1016 – 1017 cm⫺3 ), the impact ionization time would be of the order of 10 ns.16 That is consistent with the 15–20 ns delay 共relative to the 404 nm pulse兲 of the n e maximum. Third, even though the on- and off-resonance temperatures were comparable at t⫽65 ns, the 404 nm induced Na 589 nm and Li 610 nm signals were much stronger than the 407 nm case 关Figs. 3共a兲 and 3共d兲, lighter symbols兴. Assuming LTE, the difference in excited state population could not be attributed to the difference in n e . 15 The only plausible reason was a difference in the mass heated. Since the vapor plume density should be similar in both cases, a much larger volume of the plume was probably heated in the resonance case. The second observation, concerning the role of 1⫹1 photoionization 共processes 1 and 2兲, was further investigated by ablating pellets of NaHCO3 doped with lithium as the analyte. The Na-rich plume was intercepted by a dye laser beam tuned to 589 共594兲 nm for on 共off兲-resonance pumping. In this case, ionization was energetically impossible even with the absorption of a second 589 nm photon. Yet, a similar boost in the analyte 共Li 670.8 nm兲 emissions was observed. This unambiguously showed that resonant photoionization is not essential for signal enhancement. The spatial extent 共FWHM兲 of the Li 670.8 nm emissions was also estimated by scanning the plume image vertically across a 50 m horizontal slit mounted in front of the spectrograph slit. The resonant image was about 400 m tall while the offresonance one was about 130 m. This supports the third observation. The difference in heated volume may be explained as follows. An off-resonance laser pulse deposits energy in a plasma via inverse Bremsstrahlung absorption. For visible light, that absorption cross section is about 10⫺21 cm2 共Ref. 17兲. With n e of 1017 cm⫺3 , the absorption coefficient ␣ (⫽ n e ) is only 10⫺4 cm⫺1 . Hotter regions near the beam focus may absorb more light because of higher n e . So, hot spots become hotter in a positive feedback fashion leading to localized and unpredictable heating.16 In sharp contrast, a resonant laser pulse deposits energy in the host atoms extremely efficiently. The absorption cross section of the K 404 nm transition is about 2⫻10⫺16 cm2 . The number density of host atoms is about 1018 cm⫺3 共Refs. 16 and 17兲. The absorption coefficient is therefore ⬃102 cm⫺1 . Uniform deposition of light energy in an extended volume is possible because local absorption is automatically capped whenever the excited population is saturated.15 Subsequent superelastic collisions distribute that electronic energy evenly throughout the plasma plume in the form of heat. This sustains an LTE plasma at a temperature favorable for the spectrochemical analysis of sodium or lithium.18 Our observations have important practical implications. First, unlike photoionization, superelastic collisions thermalize the absorbed energy without directly generating more unwanted free electrons. A resonant excitation scheme is also simpler to devise than resonant photoionization, as borne out by our 589 nm excitation of Na. An ultraviolet 330 nm 1⫹1 scheme or less probable 660 nm 2⫹1 transitions would be required to resonantly photoionize Na.13 Second,
S. L. Lui and N. H. Cheung
resonant absorption, as opposed to inverse Bremsstrahlung heating, offers efficient and controllable energy delivery to targeted atoms. Self-capping prevents local overheating. The more extensive heating means a larger heat reservoir at the optimal temperature, leading to prolonged signals. In other words, maintaining an optimal temperature is the key to analytical sensitivity. Of course, another effective way to maintain the temperature is to reduce the heat lost from the plasma. A rarefied atmosphere would better insulate the plume and prolong the signal.19 To test the idea, we reduced the air pressure from one atmosphere to below 0.1 mbar.20 The analyte signal increased until around 350 mbar and then decreased. At lower pressure, the freer expansion and accelerated thinning of the plume caused a drop in signal and n e . A heavy inert gas, such as xenon, would confine the plume even at low pressure while hardly conducting heat away. Our study showed that 13 mbar of xenon gave the best SNR. The corresponding RELIPS spectrum is shown in Fig. 2 共trace e兲. The SNR, as defined earlier, is about 110. In summary, RELIPS delivered a superior analytical performance because it kept a larger plasma volume at the preferred temperature for a longer time. Interestingly, one may stretch the dye laser pulse to tens or even hundreds of s, such as with a flash lamp pumped device. The sustained plasma and the extended signal lifetime should enhance the SNR tremendously. RELIPS studies with a long-pulse dye laser are presently underway. This work was supported by the Research Grants Council of the University Grants Committee of Hong Kong and Faculty Research Grants from Hong Kong Baptist University. 1
For a review of LIPS, see D. A. Rusak, B. C. Castle, B. W. Smith, and J. D. Winefordner, Crit. Rev. Anal. Chem. 27, 257 共1997兲. 2 A. K. Knight, N. L. Scherbarth, D. A. Cremers, and M. J. Ferris, Appl. Spectrosc. 54, 331 共2000兲. 3 D. Anglos, K. Melesanaki, V. Zafiropulos, M. J. Gresalfi, and J. C. Miller, Appl. Spectrosc. 56, 423 共2002兲. 4 S. Y. Chan and N. H. Cheung, Anal. Chem. 72, 2087 共2000兲. 5 At lithium concentrations higher than 100 ppm, self-absorption of the 670.8 nm line became significant. 6 K. M. Lo and N. H. Cheung, Appl. Spectrosc. 56, 682 共2002兲. 7 C. W. Ng, W. F. Ho, and N. H. Cheung, Appl. Spectrosc. 51, 976 共1997兲. 8 J. D. Wu and N. H. Cheung, Appl. Spectrosc. 55, 366 共2001兲. 9 All laser fluence refers to the peak fluence of the central hottest 共⬃90% maximum兲 region. 10 According to the ICCD manufacturer, an electronic pulse width of 50 ns gave an optical width of 37 ns. 11 H. P. Gu, Q. H. Lou, N. H. Cheung, S. C. Chen, Z. Y. Wang, and P. K. Lim, Appl. Phys. B: Lasers Opt. B58, 143 共1994兲. 12 H. R. Griem, Plasma Spectroscopy 共McGraw–Hill, New York, 1964兲. 13 G. S. Hurst and M. G. Payne, Principles and Applications of Resonance Ionization Spectroscopy 共Hilger, Bristol, 1988兲. 14 R. M. Measures and P. G. Cardinal, Phys. Rev. A 23, 804 共1981兲. 15 L. St-Onge, M. Sabsabi, and P. Cielo, Spectrochim. Acta, Part B 53, 407 共1998兲. 16 C. R. Phipps and R. W. Dreyfus, in Laser Ionization Mass Analysis, edited by A. Vertes, R. Gijbels, and F. Adams 共Wiley, New York, 1993兲. 17 The ionization fraction of potassium at T⬃0.45 eV and n e ⬃1017 cm⫺3 can be estimated from the Saha equation and is about 10%. That gives 关 K兴 ⬃1018 cm⫺3 . Typical LIPS plume density is also known to be about 1018 to 1019 cm⫺3 . 18 S. F. Wong, Honors thesis, Hong Kong Baptist University, 1998. 19 Y. Iida, Spectrochim. Acta, Part B 45, 1353 共1990兲. 20 Effects of ambient gas on RELIPS will be reported elsewhere.
Downloaded 27 Dec 2002 to 158.182.14.135. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
